Abstract: A prediction method for chaotic time series of BP neural based on DE is proposed to overcome the problems such as long computing time and easy to fall into local minimum by incorporating Differential Evolution（DE） and neural network. DE is used to optimize the weights and thresholds of BP neural network, and the BP neural network is used to search for the optimal solution. The efficiency of the proposed prediction method is tested by the simulation of three typical nonlinear systems, and the precision of this algorithm is compared with BP algorithms. The simulation results show that the proposed method has better nonlinear fitting ability and higher forecasting accuracy.

Key words: chaotic time series, Back Propagation（BP） neural, Differential Evolution（DE）, prediction

摘要： 针对BP神经网络预测模型收敛速度慢和容易陷入局部极小值的缺点，将差分进化算法和神经网络结合起来，提出了一种基于差分进化算法的BP神经网络预测混沌时间序列的方法，利用差分进化算法的全局寻优能力对BP神经网络的权值和阈值进行优化，然后训练BP神经网络预测模型求得最优解，将该预测方法用到3个典型的混沌时间序列进行算法的有效性验证，并与BP算法的预测精度进行了比较，仿真结果表明该方法对混沌时间序列预测具有更好的非线性拟合能力和更高的预测准确性。

关键词: 混沌时间序列, 反向传播（BP）神经网络, 差分进化, 预测